How to feed fit feedback and return reasons back into pattern-making
Return reasons and fit comments contain precise clues about patterns and grading, but rarely reach the technical design team. A step-by-step guide to closing the loop with AI and better data.
KEY TAKEAWAYS Summary by the editors
- Return reasons and customer fit comments can tell pattern makers which styles, sizes and body areas cause problems, but only if they are captured at SKU and size level and routed to technical design.
- Language models can classify free-text fit comments into body areas and directions, such as 'tight across hips' or 'sleeves too long', turning thousands of comments into a ranked list of fit issues.
- Zalando says it combines brand measurements, return reasons, fit feedback and findings from trained fitting models to flag sizing issues, and shares return insights with brands to show whether size and fit is a recurring problem.
- Fit issues that appear in only one size usually point to grading, while issues across all sizes point to the base pattern or the fit intent.
- Changes made from return data should be validated with fit sessions before production, because return feedback describes symptoms rather than exact pattern corrections.
Feeding fit feedback into pattern-making means turning return reasons, fit ratings and customer comments into specific, measurable corrections for the next production run. The practical route is to capture return reasons at SKU and size level, use AI to classify free-text comments by body area and direction, compare the results with the graded specification, and let the technical design team decide which corrections to test in a fit session.
Why does fit feedback rarely reach pattern makers?
In most fashion companies, return data lives with e-commerce and logistics, while patterns and grading live with technical design. The two teams use different systems and different vocabularies. A customer writes “felt tight when sitting”; a pattern maker needs to know whether the seat, the rise or the waist is the problem, in which sizes, and by how much. Without a bridge between the two, return reasons are used to manage customer service rather than to improve the product.
The cost is significant. Zalando wrote in June 2026 that size and fit account for up to half of European online fashion returns, which can reach about 50% overall. In the US, the National Retail Federation estimated that online returns equalled 19.3% of online sales across all retail in 2025. Many of those returns repeat season after season on carry-over styles, because the underlying pattern or grading is never adjusted.
What fit signals are available?
| Signal | Source | What it reveals | Limitation |
|---|---|---|---|
| Structured return reason | Returns portal, store returns | Direction of misfit: too small, too large, too long | No body area |
| Free-text return comment | Returns portal | Body area and situation, for example hips, sleeves, when sitting | Unstructured, varied language |
| Fit rating and reviews | Product pages | Perceived fit relative to expectations | Self-selected respondents |
| Size exchange patterns | Order and return data | Which size customers switch to | Needs exchange tracking |
| Retailer return insights | Wholesale and marketplace partners | Recurring issues in other channels | Aggregated, may lack detail |
| Fit model sessions | Internal or partner fitting | Precise, measured observations | Limited body types and samples |
Large retailers already combine several of these. Zalando says its size flags use brand-provided measurements, customers' return history and return reasons, and findings from a dedicated team of fitting models who try on items, and that it shares return insights with brands to show whether size and fit is a recurring issue in their categories. For brands selling through such partners, these insights are an additional input to the loop.
How can AI turn comments into fit issues?
Free-text comments are where the detail sits, and they are where language models help most. A model can be instructed to read each comment and extract structured fields: the body area (shoulders, chest, waist, hips, thigh, rise, length), the direction (too tight, too loose, too long, too short) and the situation (when sitting, after washing). Aggregated by style and size, these become a ranked list of fit issues that a technical designer can act on.
- Define a fixed list of body areas and directions that matches your measurement points.
- Have the model return only those categories, plus a confidence flag, so output is consistent.
- Check a sample of classifications manually each season to catch drift or misreadings.
- Keep the original comment linked to each classification for reference.
- Exclude personal data that is not needed for the analysis.
How do you read the results as a pattern maker?
The pattern of issues across sizes is the key diagnostic. If complaints of tightness at the hips appear evenly across all sizes, the base pattern or the fit intent is likely the cause. If they appear mainly in the largest or smallest sizes, the grading rules are the more likely culprit. If complaints cluster in one production lot, the cause may be fabric shrinkage or a factory deviation rather than the pattern.
Return data describes symptoms, not corrections. A cluster of “sleeves too long” comments does not say by how many centimetres to shorten them, and customers' expectations may differ from the intended fit. That is why the loop ends in a fit session, not an automatic pattern change.
What does a closed fit feedback loop look like step by step?
- Capture: record a structured return reason for every return, linked to SKU and size, and allow an optional comment.
- Classify: use a language model to extract body area, direction and situation from comments.
- Aggregate: calculate return rates and issue counts by style, size and production lot.
- Prioritise: focus on carry-over and high-volume styles where a correction will affect future orders.
- Diagnose: technical design compares issues with the graded specification and fabric data.
- Test: make the correction and validate it in a fit session on relevant sizes.
- Update: change the pattern, grading or fit description, and record the change against the style.
- Measure: compare size-related return rates before and after the change in the next season.
Who should own the loop?
The loop works best with a named owner in technical design or product development, supported by an analyst from e-commerce. Monthly reviews of the top fit issues, with decisions recorded against each style, keep the process from fading after the first season. Brands that sell through wholesale partners should ask key accounts and marketplaces for aggregated return reasons, since those channels may see different customers and problems.
What are the limits?
Return data is skewed towards customers who bother to explain, and online returns reflect expectations set by imagery and descriptions as much as by fit. Some issues, such as a fabric that feels scratchy, are not pattern problems at all. AI makes it feasible to read thousands of comments, but the judgement about what to change and by how much remains with pattern makers and fit technicians.
Frequently asked questions
How can return data improve garment fit?
Return reasons and comments, linked to style and size, show which garments and body areas cause fit problems. Technical designers can use this to adjust patterns or grading, then validate changes in fit sessions before the next production run.
Can AI analyse customer fit comments?
Yes. Language models can classify free-text comments into body areas, directions such as too tight or too long, and situations. Aggregated by style and size, the results show where fit problems cluster, though a sample should be checked manually.
How do you know if a fit problem is grading or the base pattern?
If an issue appears evenly across all sizes, the base pattern or fit intent is the likely cause. If it is concentrated in the smallest or largest sizes, grading rules are more likely. Issues in a single production lot may point to fabric or factory deviations.
Who should own fit feedback in a fashion company?
Ideally technical design or product development, with analytical support from e-commerce. Clear ownership and regular reviews prevent return insights from staying with customer service and logistics.
One edition every weekday morning. Read in five minutes. Free for industry professionals.
